高速公路路元快速检测方法

发布时间：2018-04-05 13:37

本文选题：高速公路检测　切入点：路面管理系统　出处：《湖北工业大学》2017年硕士论文

【摘要】：随着我国的交通运输行业地不断发展,公路不断地建设,公路里程数也不断地增加。目前,我国总公路里程数已经超过美国成为世界第一。随之而来的是我国公路的大规模养护,为了完成公路大规模的管理养护工作,路面管理养护系统应运而生,该系统使用线性里程桩号的方法定位路面病害,并记录路面病害的发展与维修历史。然而,线性里程桩号定位方法存在着路面管理单元的划分过大,不能精确的定位路面病害的问题,无法实现养护路面的精细化管理。为了解决路面病害不能精确定位的问题,以及里程桩号的断链问题,路元概念被提出并得到了较好应用。路元是最小路面管理单元,该理念的提出可以科学有效地解决里程桩号定位不准确的问题。但是路元以其分布密度大、里程长、面积小等特点对路元信息的采集与识别工作提出了挑战,目前路元信息的自动化采集与识别的研究尚未有人涉及。本文针对路元信息提出了一个快速高效的采集方案,并针对采集到的路元图像,提出了基于双阈值检测的路元标志定位方法,该方法是利用图像的灰度直方图将路元图像同时进行高、低两个阈值的处理,并将处理后的图像进行叠加,得到路元的定位图像。该方法中高低阈值的确定是其关键部分。通过实验证明,基于双阈值的路元定位方法定位准确率高于传统的基于边缘检测的定位方法。最后,针对路元标志中字符性变大,干扰多等特点,提出了改进的高斯滤波方法,能够有效的滤除杂质并增强字符边缘信息,并利用BP神经网络完成对的路元字符的识别,该方法能够有效的识别部分缺损、污染的字符。本文选取1000张在某省高速公路上采集到的路元图像,并采用上述方法对路元标志进行定位和识别,其定位精度达到95%,识别精度能够达到92.5%。通过实验证明,本文提出的定位识别方法高效可行,系统工作运行安全可靠,为高速公路的快速检测提供了有效的方法和手段。
[Abstract]:With the development of transportation industry and the construction of highway, the mileage of highway is increasing.At present, China's total road mileage has surpassed the United States to become the first in the world.The following is the large-scale maintenance of highway in our country. In order to complete the large-scale management and maintenance of highway, the pavement management and maintenance system emerges as the times require, and the system uses the method of linear mileage pile number to locate the pavement disease.The development and maintenance history of pavement diseases were recorded.However, the linear mileage pile number location method has the problem that the pavement management unit is too large to accurately locate the pavement disease, and the fine management of the maintenance pavement can not be realized.In order to solve the problem that the road diseases can not be located accurately and the mileage pile number is broken, the concept of road element has been put forward and applied well.The road element is the minimum pavement management unit, which can solve the problem of inaccurate location of mileage pile number scientifically and effectively.However, because of its large distribution density, long mileage and small area, road elements have challenged the acquisition and recognition of road element information. At present, the research on automatic collection and recognition of road element information has not been involved.In this paper, a fast and efficient method for road element information acquisition is proposed, and a road element sign location method based on dual threshold detection is proposed for the collected road element image.This method uses the gray histogram of the image to process the road element image at the same time with two thresholds: high and low, and the processed image is superposed to obtain the location image of the road element.The determination of high and low threshold is the key part of this method.The experimental results show that the accuracy of road element localization based on double threshold is higher than that of traditional edge detection.Finally, in view of the character character becoming bigger and the interference more in the road element sign, the improved Gao Si filter method is put forward, which can effectively filter the impurity and enhance the character edge information, and use the BP neural network to complete the recognition of the road element character.This method can effectively identify some defective and contaminated characters.In this paper, 1000 road element images collected on highway in a province are selected, and the above methods are used to locate and identify the road element signs. The positioning accuracy is 95% and the recognition accuracy can reach 92.5%.The experimental results show that the proposed method is efficient and feasible, and the system is safe and reliable in operation, which provides an effective method and means for the rapid detection of freeway.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U418.4;TP391.41

【参考文献】